Cert Pepper

CertPepper

IT LEARNS WHAT YOU DON'T KNOW.

Adaptive CLI study tool for IT certification exams — FSRS spaced repetition meets AI-powered explanations.

Quick Start · Why CertPepper? · Features · MCP Integration · Content Format · Contributing

Security+ in 10 Days →

Quick Start

# Clone and install
git clone https://github.com/cert-pepper/cert-pepper.git && cd cert-pepper
cp .env.example .env          # add ANTHROPIC_API_KEY for AI explanations
uv sync

# Set up a database and start studying
uv run cert-pepper db init
uv run cert-pepper ingest     # loads the bundled Security+ content
uv run cert-pepper study

Or open the repo in Claude Code and ask it to generate a question bank for any exam:

> Set up a standard question bank for the CISSP exam

That request should map to the content MCP server's create_question_bank workflow. It creates or reuses a DB-backed question bank; it does not create a new examples/ content pack unless you explicitly ask for repository content files.

See [MCP Integration](#mcp-integration) to enable the servers.

Why CertPepper?

Most study tools — Anki, Quizlet, flashcard apps — treat every question the same. CertPepper adapts:

• FSRS-4.5 spaced repetition schedules reviews at optimal intervals based on your recall history — the same core algorithm behind Anki, tuned for exam prep.
• Bayesian Knowledge Tracing estimates mastery per domain and steers new questions toward your weakest areas, weighted by how much each domain counts on the real exam.
• AI explanations break down wrong answers with domain-specific context via Claude (MCP sampling in Claude Code, or the Anthropic API in CLI mode).
• Domain-weighted scoring predicts your exam score using actual exam weights, with a conservative prior for unseen material.
• MCP integration — study inside Claude Code with three servers for content, study sessions, and analytics.

Features

| Command | What it does | |---------|-------------| | cert-pepper study | Adaptive study session — picks questions based on your weak spots | | cert-pepper study --domain 4 --count 15 | Target a specific domain | | cert-pepper study --new-questions | Focus on unseen material | | cert-pepper exam | 90-question timed mock exam with countdown | | cert-pepper progress | Dashboard: domain accuracy, predicted score, pass probability, study streak | | cert-pepper goal set --exam-date 2026-03-09 | Set exam date, get a paced study calendar | | cert-pepper pregenerate | Batch-generate AI explanations (requires API key) | | cert-pepper db init | Create/reset the SQLite database | | cert-pepper ingest | Parse markdown content into the DB | | cert-pepper upgrade | Apply schema migrations and refresh content |

> No guarantees. Predicted scores and pass probabilities are estimates based on practice performance — not predictions of real exam outcomes.

MCP Integration

Three STDIO MCP servers are registered in .mcp.json:

| Server | Purpose | |--------|---------| | cert-pepper-study | Start sessions, submit answers, get due cards | | cert-pepper-content | Search questions, get explanations, look up acronyms | | cert-pepper-analytics | Predict score, find weak areas, study recommendations |

Enable them in Claude Code by adding enableAllProjectMcpServers: true to .claude/settings.local.json.

In Claude Code, explanations use MCP sampling — no ANTHROPIC_API_KEY needed.

For exam setup, the content server exposes two equivalent tools:

• create_question_bank(exam_name, size=None) — preferred user-facing entrypoint
• setup_exam(exam_name, size=None) — lower-level alias with the same behavior

Both tools accept size tiers lite, standard, and heavy. They create or reuse DB-backed question banks rather than writing content under examples/.

Adding Your Own Exam

1. Create a content directory in CertPepper's [markdown format](docs/content-format.md).
2. Set CONTENT_ROOT=/path/to/your/content in .env.
3. Run cert-pepper db init and cert-pepper ingest.

The examples/security-plus/ directory is a complete reference implementation with 228 questions, 135 flashcards, and 262 acronyms.

Configuration

| Variable | Default | Description | |----------|---------|-------------| | DB_PATH | ./cert_pepper.db | SQLite database path | | CONTENT_ROOT | ./examples/security-plus | Root of your exam content | | ANTHROPIC_API_KEY | — | Required for CLI AI explanations (study, pregenerate). Not needed in Claude Code. | | HAIKU_MODEL | claude-haiku-4-5-20251001 | Model for AI explanations | | SONNET_MODEL | claude-sonnet-4-6 | Model for MCP sampling | | DEFAULT_SESSION_SIZE | 10 | Questions per study session | | MASTERY_THRESHOLD | 0.85 | BKT mastery cutoff |

Repository Layout

cert-pepper/
├── cert_pepper/          — Python source (CLI, MCP servers, algorithms)
│   ├── cli/              — Typer commands
│   ├── mcp/              — Three FastMCP STDIO servers
│   ├── engine/           — FSRS, BKT, selector, scorer (pure Python)
│   ├── ingestion/        — Markdown parsers
│   ├── ai/               — Anthropic client + explainer
│   └── db/               — SQLAlchemy async engine + schema
├── tests/                — pytest suite
├── examples/
│   └── security-plus/    — Security+ SY0-701 exam content
└── docs/
    ├── security-plus-in-10-days.md — 10-day Security+ study guide
    └── content-format.md — Format spec for your own exam content

Worked Example

The examples/security-plus/ directory contains a complete Security+ SY0-701 exam prep set:

• 5 domains of notes
• 135 flashcards
• 228 practice questions across all 5 domains
• 262 acronyms

See [Security+ in 10 Days](docs/security-plus-in-10-days.md) for a step-by-step guide.

Contributing

See [CONTRIBUTING.md](CONTRIBUTING.md) for setup instructions, workflow, and PR guidelines.

License

[MIT](LICENSE)

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: cert-pepper
• Source: cert-pepper/cert-pepper
• License: MIT

Install and usage instructions live in the source repository linked above.